Rapid screening method of arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy

Abstract

The application discloses a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy, and relates to the field. The arsenic-specific quantum dot labeling probe is composed of a first quantum dot nanocrystal core and an arsenic-specific recognition group. The mercury-specific quantum dot labeling probe is composed of a second quantum dot nanocrystal core and a mercury-specific recognition group. The application adopts a dual-mode detection mechanism of quantum dot labeling to enhance terahertz fingerprint identification, discards traditional biological living body metabolic response detection modes and liquid-phase fluorescence detection modes, realizes directional combination of specific quantum dot probes and arsenic and mercury elements, forms a compound with characteristic terahertz vibration signals, greatly improves the detection sensitivity and specificity of heavy metal elements by terahertz technology, and realizes accurate differentiation and synchronous detection of the two target elements of arsenic and mercury.

Description

Technical Field

[0001] This invention relates to the field of food safety testing technology, and in particular to a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy. Background Technology

[0002] Algae powder, a common functional food ingredient, is rich in protein, polysaccharides, and various active nutrients, and has a wide range of applications. During the growth of microalgae, they readily accumulate heavy metals such as arsenic and mercury from water and the environment. Excessive levels of heavy metals can seriously affect the safety of algae powder products. Therefore, accurate and rapid detection of arsenic and mercury content in algae powder is a crucial step in ensuring its food safety.

[0003] Currently, the mainstream detection methods for heavy metals in algal powder are inductively coupled plasma mass spectrometry (ICP-MS) and inductively coupled plasma atomic emission spectrometry (ICP-AES). These methods require complex pretreatment processes such as sample digestion, resulting in long overall detection cycles, expensive equipment, and high operator skill requirements, making them unsuitable for rapid screening of large batches of samples. Furthermore, the complex matrix composition of algal powder, with its large amount of endogenous organic components, can cause matrix interference in traditional detection systems, reducing detection stability and accuracy.

[0004] Existing terahertz time-domain spectroscopy can be used for heavy metal detection in biological samples. The core of this technology relies on the changes in the spectral response of endogenous substances in organisms affected by heavy metals. However, this detection method depends on the metabolic reaction of living organisms, which is time-consuming. Furthermore, terahertz spectroscopy has low sensitivity to the identification of heavy metal elements themselves, and can only achieve a broad qualitative judgment of heavy metals. It cannot achieve the simultaneous and specific differentiation of arsenic and mercury, making it difficult to adapt to the precise screening scenario of trace heavy metals in solid algal powder.

[0005] Quantum dot fluorescence detection technology has the advantages of high sensitivity and strong specificity in heavy metal detection. Existing related detection schemes are mostly based on liquid phase detection systems, which require solid samples to be dissolved and buffer solutions to be prepared. They cannot be directly used for in-situ detection of solid algal powder samples. The sample pretreatment process is cumbersome, the detection efficiency is low, and rapid batch screening cannot be achieved.

[0006] In summary, there is currently a lack of a rapid screening technology that is compatible with solid algae powder matrix, can simultaneously detect arsenic and mercury, has a fast detection speed, and strong anti-interference ability. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy. The technical solution adopted is as follows:

[0008] A rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy includes the following steps:

[0009] Step 1: Prepare a standardized algae powder test film from the algae powder sample;

[0010] Step 2: Prepare arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes respectively; the arsenic-specific quantum dot labeled probe is composed of a first quantum dot nanocrystal core and an arsenic-specific recognition group; the mercury-specific quantum dot labeled probe is composed of a second quantum dot nanocrystal core and a mercury-specific recognition group.

[0011] Step 3: Arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes are directionally labeled and incubated with the algae powder test film, so that the quantum dot labeled probes specifically bind to arsenic and mercury in the algae powder sample. After washing and drying, the quantum dot labeled algae powder test film is obtained.

[0012] Step 4: Perform dual-modal synchronous data acquisition on the quantum dot-labeled algal powder test film on the same measurement platform, including: acquiring fluorescence microscopic images of the quantum dot-labeled probe in quantum microscopic imaging mode to locate the quantum dot-labeled enriched micro-regions; and acquiring terahertz time-domain spectral signals of each enriched micro-region in terahertz time-domain spectral mode, guided by the located enriched micro-regions.

[0013] Step 5: Perform multimodal fusion analysis on the collected dual-modal data, including: extracting the differential terahertz absorption spectrum of the quantum dot labeled region relative to the unlabeled region, constructing a joint feature vector that integrates terahertz spectral features and quantum micro-fluorescence features, and simultaneously inverting arsenic and mercury concentrations based on a causal inference multi-task learning model.

[0014] Step 6: Compare the obtained arsenic and mercury concentrations with the preset limit thresholds and output the screening judgment results.

[0015] Optionally, step 1 includes the following sub-steps:

[0016] Step 11: Weigh 100±0.1mg of algae powder sample, add high-purity anhydrous ethanol as a dispersion medium, grind into a uniform paste, and then uniformly coat it onto a high-resistivity silicon wafer substrate.

[0017] Step 12: Use a spin coater to spin coat algae powder at a speed of 700rpm-900rpm to form a film with a thickness of 100±20μm;

[0018] Step 13: Dry the film under a nitrogen atmosphere at 38℃-42℃ for 15-18 minutes to obtain the algae powder test film; the thickness of the high-resistivity silicon wafer substrate is 500±10μm, the resistivity is greater than 10kΩ·cm, and the transmittance in the 0.1-4.0THz frequency band is not less than 60%.

[0019] Optionally, the preparation method of the arsenic-specific quantum dot-labeled probe in step 2 is as follows: CdSe / ZnS core-shell quantum dots with an emission wavelength of 520±5 nm and a quantum yield greater than 60% are selected as the core of the first quantum dot nanocrystal. Glutathione molecules are modified on its surface via ligand exchange. After activation with 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide, an arsenic-specific recognition peptide is coupled. The amino acid sequence of the arsenic-specific recognition peptide is Cys-Gly-Gly-Cys-Gly-Gly. After ultrafiltration purification, the arsenic-specific quantum dot-labeled probe is obtained. The glutathione bonded to the surface of the algal powder test film probe is in situ reduced to As(V) to As under incubation conditions. 3+ Then, As is chelated via peptide dithiol coordination. 3+ Trivalent methylarsenic has no specific binding ability to pentavalent methylarsenic or arsenose.

[0020] Optionally, the preparation method of the mercury-specific quantum dot-labeled probe in step 2 is as follows: CdSe / ZnS core-shell structured quantum dots with an emission wavelength of 620±5nm and a quantum yield greater than 55% are selected as the core of the second quantum dot nanocrystal. They are converted into water-soluble quantum dots modified with dihydrolipoic acid by ligand exchange method. Subsequently, a thymine-rich single-stranded DNA sequence (5'-SH-(CH2)6-TTTTTTTTTTTTTTTT-3') is grafted onto its surface via Au-S bonds, and Hg is added. 2+ Hybridization was performed using the aptamer sequence (5'-TTCTTTCTTCCCCTTGTTTGTT-3') based on T-Hg. 2+ -T coordination chemistry mechanism enables specific recognition of mercury ions.

[0021] Optionally, step 3 includes the following sub-steps:

[0022] Step 31: Mix arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes at a molar ratio of 1:1, dilute with phosphate buffer to a total quantum dot concentration of 10 nM, and add 0.05% Tween-20 surfactant to obtain a dual-channel quantum dot labeled probe mixture.

[0023] Step 32: 100 μL of the dual-channel quantum dot labeled probe mixture is uniformly dropped onto the surface of the algae powder test film and incubated in a humidity-saturated sealed container at 36-40°C in the dark for 28-35 minutes.

[0024] Step 33: After incubation, rinse the film surface with phosphate buffer to remove unbound probes, and dry it under nitrogen flow to obtain a quantum dot labeled algae powder test film.

[0025] Optionally, the method for locating the enriched micro-regions in step 4 is:

[0026] An arsenic-specific quantum dot-labeled probe was excited with a 405 nm laser, and fluorescence microscopy images were acquired in the 520 nm channel. A mercury-specific quantum dot-labeled probe was excited with a 532 nm laser, and fluorescence microscopy images were acquired in the 620 nm channel. The imaging field of view was 500 μm × 500 μm. Pixel-level superposition analysis was performed on the two fluorescence images. An adaptive threshold segmentation algorithm was used to identify pixel regions with fluorescence intensity higher than three times the background mean, which were marked as quantum dot-labeled enriched microregions. The center coordinates and corresponding fluorescence intensity values ​​of each enriched microregion were recorded. The quantum dot-labeled enriched microregions represent the potential arsenic / mercury presence areas located by the specific binding of the quantum dot probe.

[0027] Optionally, the method for acquiring the terahertz time-domain spectral signal in step 4 is as follows: based on the determined coordinates of the enriched micro-regions, the electric sample stage is controlled to move each enriched micro-region sequentially to the focal position of the terahertz beam, and the terahertz time-domain spectral signal is acquired in transmission mode; the terahertz frequency range is 0.1THz-4.0THz, the terahertz spot diameter is 100μm-150μm, and each spectrum is acquired by 100-150 time-domain scans and the average value is taken; at the same time, the terahertz time-domain spectrum of the blank area without fluorescent label is acquired as a reference signal; the ambient temperature is controlled at 23±1 degrees Celsius and the relative humidity is below 5%.

[0028] Optionally, the multimodal fusion analysis in step 5 includes the following sub-steps:

[0029] Step 51: Perform a fast Fourier transform on the terahertz time-domain spectral signal and the reference signal acquired in the previous step to calculate the terahertz absorption coefficient α(ν) and refractive index n(ν).

[0030] Step 52: For each quantum dot-labeled enriched micro-region i, calculate the terahertz absorption coefficient of the quantum dot-labeled region. Absorption coefficient relative to the unlabeled region of the same sample Differential absorption spectrum To eliminate the contribution of algal powder matrix background absorption;

[0031] Step 53, construct the joint feature vector , ;in These are discrete frequency points selected at equal intervals within the range of 0.1THz to 4.0THz, where N is the total number. It is the normalized fluorescence intensity of the arsenic-specific quantum dot in the i-th enriched micro-region; It is the normalized fluorescence intensity of the mercury-specific quantum dot in the i-th enriched micro-region; It is the area of ​​the i-th enriched micro-region.

[0032] Optionally, the multi-task learning model based on causal inference in step 5 adopts a causal multi-task learning architecture, specifically including:

[0033] Construct a causal graph structure G=(V,E), where the node set V contains terahertz spectral feature nodes. Quantum dot fluorescence characteristic nodes Algal powder matrix confusion factor node Z, arsenic concentration node and mercury concentration nodes The edge set E is composed of causal relationships determined by domain knowledge;

[0034] A dual-channel convolutional neural network is used as the encoder for terahertz spectral features and quantum dot fluorescence features, and a fully connected regression head is used as the predictor for arsenic and mercury concentrations. A causal regularization term is introduced to eliminate the confounding effect of algal powder matrix confusion factors and ensure cross-batch generalization ability.

[0035] Optionally, step 6 can be further described as follows: The predicted arsenic concentration obtained from step 5 can be inverted... and predicted mercury concentration The values ​​are compared with the set limits for arsenic and mercury in algae and their products; the 95% confidence interval of the model prediction is estimated using the Bayesian dropout method; if the predicted concentration exceeds the corresponding limit, it is judged as exceeding the standard, otherwise it is judged as qualified, and the judgment conclusion, the predicted concentration value and its confidence interval are output together to generate a standardized test report.

[0036] In summary, the present invention has at least one of the following beneficial technical effects:

[0037] This invention provides a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy. It employs a dual-modal detection mechanism using quantum dot labeling to enhance terahertz fingerprint recognition, abandoning traditional biological in vivo metabolic response detection and liquid-phase fluorescence detection modes. Through the targeted binding of specific quantum dot probes with arsenic and mercury, a complex with characteristic terahertz vibrational signals is formed, significantly improving the sensitivity and specificity of terahertz technology for heavy metal elements, achieving accurate differentiation and simultaneous detection of arsenic and mercury. The arsenic-targeting probe relies on glutathione in-situ reduction to achieve As(V) conversion detection, enabling precise screening only for inorganic arsenic and trivalent methylarsenic in algal powder. It is not suitable for the quantitative detection of pentavalent methylarsenic and arsenic sugars, but can meet the need for rapid initial screening of inorganic arsenic risks in algal powder.

[0038] By leveraging quantum microscopy to achieve precise localization of heavy metal-enriched micro-regions, and guiding terahertz spectroscopy to perform focused detection of these micro-regions, the system effectively avoids background interference from complex algal powder matrices, improving the signal-to-noise ratio and detection accuracy. Simultaneously, through a causal inference multi-task learning algorithm, the system eliminates the influence of matrix contamination factors, optimizes the model's generalization ability across batches, and addresses the issues of poor interpretability and insufficient stability in traditional machine learning-based spectral detection.

[0039] The pretreatment process for algae powder samples has been simplified, eliminating the need for high-temperature digestion and complex liquid phase preparation. The detection process is simple and efficient, significantly shortening the overall detection cycle. It can achieve in-situ rapid screening of trace arsenic and mercury in solid algae powder samples, balancing detection accuracy, efficiency, and stability. It can meet the application needs of various scenarios such as batch self-inspection by food production enterprises and rapid sampling inspection by market supervision. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to the present invention.

[0041] Figure 2 This is a schematic diagram of the test report result page according to a specific embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the test report data quality and judgment explanation page according to a specific embodiment of the present invention. Detailed Implementation

[0043] The present invention will be further described in detail below with reference to the accompanying drawings.

[0044] This invention discloses a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy.

[0045] Reference Figures 1-3 Example 1, a rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy, includes the following steps:

[0046] Step 1: Prepare a standardized algae powder test film from the algae powder sample;

[0047] Step 2: Prepare arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes respectively; the arsenic-specific quantum dot labeled probe is composed of a first quantum dot nanocrystal core and an arsenic-specific recognition group; the mercury-specific quantum dot labeled probe is composed of a second quantum dot nanocrystal core and a mercury-specific recognition group.

[0048] Step 3: Arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes are directionally labeled and incubated with the algae powder test film, so that the quantum dot labeled probes specifically bind to arsenic and mercury in the algae powder sample. After washing and drying, the quantum dot labeled algae powder test film is obtained.

[0049] Step 4: Perform dual-modal synchronous data acquisition on the quantum dot-labeled algal powder test film on the same measurement platform, including: acquiring fluorescence microscopic images of the quantum dot-labeled probe in quantum microscopic imaging mode to locate the quantum dot-labeled enriched micro-regions; and acquiring terahertz time-domain spectral signals of each enriched micro-region in terahertz time-domain spectral mode, guided by the located enriched micro-regions.

[0050] Step 5: Perform multimodal fusion analysis on the collected dual-modal data, including: extracting the differential terahertz absorption spectrum of the quantum dot labeled region relative to the unlabeled region, constructing a joint feature vector that integrates terahertz spectral features and quantum micro-fluorescence features, and simultaneously inverting arsenic and mercury concentrations based on a causal inference multi-task learning model.

[0051] Step 6: Compare the obtained arsenic and mercury concentrations with the preset limit thresholds and output the screening judgment results.

[0052] By employing the above technical solution, quantum dot probes are used as a signal conversion bridge connecting terahertz spectroscopy and heavy metal elements. In existing technologies, heavy metal ions such as arsenic and mercury lack characteristic absorption peaks in the terahertz band and cannot be directly detected. This invention utilizes the specific binding of quantum dot-labeled probes with arsenic and mercury to form quantum dot-heavy metal complexes, which generate detectable collective vibrational mode changes in the terahertz band, thus endowing arsenic and mercury with indirect fingerprint characteristics that can be identified by terahertz spectroscopy. Simultaneously, quantum microscopy provides spatial distribution information of probe binding, guiding the terahertz spectrum to precisely focus on the positively labeled region for acquisition, reducing the terahertz detection area from the entire sample film to a micrometer-level enrichment region, making the effective signal significantly stand out from the matrix background. Finally, the dual-modal data are fused through a causal inference model, which explicitly models the algal powder matrix background as a confounding factor, severing the spurious statistical correlation between it and the predicted target, thereby outputting robust predicted values ​​unaffected by batch-to-batch matrix differences.

[0053] Example 2, step 1 includes the following sub-steps:

[0054] Step 11: Weigh 100±0.1mg of algae powder sample, add high-purity anhydrous ethanol as a dispersion medium, grind into a uniform paste, and then uniformly coat it onto a high-resistivity silicon wafer substrate.

[0055] Step 12: Use a spin coater to spin coat algae powder at a speed of 700rpm-900rpm to form a film with a thickness of 100±20μm;

[0056] Step 13: Dry the film under a nitrogen atmosphere at 38℃-42℃ for 15-18 minutes to obtain the algae powder test film; the thickness of the high-resistivity silicon wafer substrate is 500±10μm, the resistivity is greater than 10kΩ·cm, and the transmittance in the 0.1-4.0THz frequency band is not less than 60%.

[0057] By adopting the above technical solution, three key parameters for thin film preparation were limited, based on the high sensitivity of terahertz time-domain spectroscopy to sample morphology. The high-resistivity silicon substrate exhibits low absorption and high transmittance in the 0.1THz to 4.0THz frequency range. Its resistivity is greater than 10kΩ·cm to ensure that the concentration of free carriers inside the substrate is extremely low under the influence of a terahertz electric field, preventing significant absorption attenuation of the transmitted terahertz signal. If the resistivity is insufficient, carrier absorption will overwhelm the sample's own weak spectral characteristics. A film thickness of 100μm ± 20μm is an experimentally optimized compromise: too thin a thickness leads to insufficient optical path for the interaction between the terahertz wave and the algal powder sample, resulting in a weak absorption signal and a reduced signal-to-noise ratio; too thick a thickness causes a sharp attenuation of the terahertz signal intensity due to the cumulative absorption of a large amount of organic matter in the algal powder, especially the complete cutoff of high-frequency components, leading to a narrowing of the effective spectral range. The spin coating speed of 700 rpm to 900 rpm and the drying temperature of 38°C to 42°C together determine the density and surface smoothness of the film. A smooth surface can reduce the diffuse reflection loss of terahertz waves at the sample interface and ensure the fidelity of the signal waveform in the transmission mode.

[0058] Example 3, the preparation method of the arsenic-specific quantum dot labeled probe in step 2 is as follows: CdSe / ZnS core-shell structured quantum dots with an emission wavelength of 520±5nm and a quantum yield greater than 60% are selected as the core of the first quantum dot nanocrystal. Glutathione molecules are modified on its surface by ligand exchange method. After activation by 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide, an arsenic-specific recognition peptide is coupled. The amino acid sequence of the arsenic-specific recognition peptide is Cys-Gly-Gly-Cys-Gly-Gly. After ultrafiltration purification, the arsenic-specific quantum dot labeled probe is obtained. The glutathione bonded to the surface of the algal powder test film probe is reduced in situ to As(V) under incubation conditions. 3+ Then, As is chelated via peptide dithiol coordination. 3+ Trivalent methylarsenic has no specific binding ability to pentavalent methylarsenic or arsenose.

[0059] The arsenic-specific recognition peptide sequence Cys-Gly-Gly-Cys-Gly-Gly-Gly, designed using the above technical solution, is based on the strong affinity of arsenic ions for the thiol groups in cysteine ​​residues. This peptide sequence contains two cysteine ​​(Cys) residues, providing two pairs of thiol groups to coordinate and chelate with trivalent arsenic ions, forming stable As-S coordination bonds. The glycine (Gly) residues provide a flexible peptide backbone, allowing the two Cys residues to be spatially aligned to their optimal positions for arsenic ion binding, reducing the impact of steric hindrance on binding efficiency. The first quantum dot uses a CdSe / ZnS core-shell structure with an emission wavelength of 520 nm. The wide bandgap ZnS shell effectively confines the electron-hole pairs of the CdSe core within the core, reducing nonradiative recombination caused by surface defect states and ensuring a quantum yield greater than 60%, providing a high-brightness fluorescence signal for subsequent quantum microscopy imaging. The ligand exchange method replaces the original hydrophobic trioctylphosphine oxide ligands on the quantum dot surface with thiol-containing glutathione, achieving efficient transfer of quantum dots from the organic phase to the aqueous phase and providing carboxyl active sites for subsequent covalent coupling. The combined use of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide converts the carboxyl group into a stable NHS ester, improving coupling efficiency with primary amine groups on the peptide chain and avoiding self-crosslinking side reactions of the peptide chain.

[0060] Meanwhile, glutathione modified on the quantum dot surface acts as an in-situ reducing agent, which, under incubation buffer conditions at 36-40℃, can reduce As(V) present in the algal powder matrix to As. 3+ The chelation was then completed; pentavalent methylarsenic DMA(V), MMA(V) and arsenose have stable chemical structures and cannot be captured and bound by this probe.

[0061] Example 4, the preparation method of mercury-specific quantum dot labeled probe in step 2 is as follows: CdSe / ZnS core-shell structured quantum dots with an emission wavelength of 620±5nm and a quantum yield greater than 55% are selected as the core of the second quantum dot nanocrystal. They are converted into water-soluble quantum dots modified with dihydrolipoic acid by ligand exchange method. Then, a thymine-rich single-stranded DNA sequence (5'-SH-(CH2)6-TTTTTTTTTTTTTTTT-3') is grafted onto its surface through Au-S bonds, and Hg is added. 2+ Hybridization was performed using the aptamer sequence (5'-TTCTTTCTTCCCCTTGTTTGTT-3') based on T-Hg. 2+ The -T coordination chemical mechanism enables the specific recognition of mercury ions. This probe can only detect inorganic divalent mercury and small molecule methylmercury; large molecule organic mercury cannot be captured and detected, so the measured mercury content is not equivalent to the total mercury in the sample.

[0062] By employing the above technical solution, mercury ions can specifically intercalate between two thymine residues, replacing the imine proton to form an N-Hg-N coordinate bond, thereby transforming the TT mismatched base pair from a repulsive state to a stable T-Hg state. 2+ -T base pairs. The thymine-rich single-stranded DNA sequence contains 15 consecutive thymine bases, providing ample high-affinity binding sites for mercury ions and ensuring efficient capture of trace amounts of mercury ions. The added aptamer sequence hybridizes complementary to a portion of the DNA backbone, enabling T-Hg binding in the presence of mercury ions. 2+ -T coordination induces the DNA backbone to fold into a hairpin structure, causing significant changes in the surface charge density and dielectric microenvironment of the quantum dots. This conformational change amplifies the optical response of the quantum dots to mercury ion binding. The second quantum dot uses a CdSe / ZnS core-shell structure with an emission wavelength of 620 nm, separated from the first quantum dot's 520 nm emission wavelength by 100 nm. The fluorescence emission peaks of the two are completely separated, enabling crosstalk-free dual-color imaging through two independent detection channels. Dihydrolipoic acid, as a bidentate thiol ligand, forms a more stable chelate coordination with zinc atoms on the quantum dot surface. Its carboxyl terminus provides good water solubility, while the terminal thiol group is indirectly connected to the thiol group modified at the 5' end of the DNA via an Au-S bond, ensuring the directional and orderly assembly of DNA on the quantum dot surface.

[0063] Example 5, step 3 includes the following sub-steps:

[0064] Step 31: Mix arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes at a molar ratio of 1:1, dilute with phosphate buffer to a total quantum dot concentration of 10 nM, and add 0.05% Tween-20 surfactant to obtain a dual-channel quantum dot labeled probe mixture.

[0065] Step 32: 100 μL of the dual-channel quantum dot labeled probe mixture is uniformly dropped onto the surface of the algae powder test film and incubated in a humidity-saturated sealed container at 36-40°C in the dark for 28-35 minutes.

[0066] Step 33: After incubation, rinse the film surface with phosphate buffer to remove unbound probes, and dry it under nitrogen flow to obtain a quantum dot labeled algae powder test film.

[0067] By adopting the above technical solution, a total quantum dot concentration of 10 nM is the optimal value determined by saturation binding experiments: concentrations below this value lead to limited labeling reaction kinetics, incomplete binding, and weak fluorescence signals in positive micro-regions; concentrations above this value increase non-specific physical adsorption, resulting in increased residual background fluorescence after washing and reduced ability of subsequent adaptive threshold segmentation to distinguish true positive micro-regions. The 0.05% (v / v) Tween-20 surfactant reduces the surface tension of the dual-channel quantum dot-labeled probe mixture, improving the spreading and wetting of the probe solution on the hydrophobic algal powder surface, allowing probe molecules to uniformly penetrate the film through the pore network between algal powder particles and reach the binding site. The incubation temperature of 36℃ to 40℃ and the saturated humidity environment maintain the film sample in a moderately swollen state. In a slightly humid environment, the molecular chain activity of polysaccharides and protein matrices in the algal powder is enhanced, increasing the effective pore size and reducing the steric hindrance of probe diffusion within the matrix. The incubation time of 28 to 35 min is the equilibrium time determined based on real-time fluorescence monitoring, at which point the fluorescence intensity reaches a plateau value, and further extending the time no longer significantly increases the labeling amount.

[0068] Example 6, the method for locating the enriched micro-regions in step 4 is:

[0069] An arsenic-specific quantum dot-labeled probe was excited with a 405 nm laser, and fluorescence microscopy images were acquired in the 520 nm channel. A mercury-specific quantum dot-labeled probe was excited with a 532 nm laser, and fluorescence microscopy images were acquired in the 620 nm channel. The imaging field of view was 500 μm × 500 μm. Pixel-level superposition analysis was performed on the two fluorescence images. An adaptive threshold segmentation algorithm was used to identify pixel regions with fluorescence intensity higher than three times the background mean, which were marked as quantum dot-labeled enriched microregions. The center coordinates and corresponding fluorescence intensity values ​​of each enriched microregion were recorded. The quantum dot-labeled enriched microregions represent the potential arsenic / mercury presence areas located by the specific binding of the quantum dot probe.

[0070] By employing the above technical solution, the quantum microscopy imaging module uses dual-wavelength excitation at 405nm and 532nm because these two excitation wavelengths are located near the absorption peaks of the two quantum dots, enabling independent and efficient excitation of arsenic-specific and mercury-specific quantum dot-labeled probes. The 520nm and 620nm channels respectively collect the characteristic fluorescence emission of the two probes. The ratio of the fluorescence intensity of the two channels can also preliminarily determine the relative content trend of arsenic and mercury in the sample. The adaptive threshold segmentation algorithm sets the threshold to the background mean plus three standard deviations. Its statistical significance lies in the fact that if the pixel fluorescence intensity follows a Gaussian distribution, the confidence interval of the background mean plus three standard deviations corresponds to approximately 99.7% confidence. Pixels exceeding this threshold have a very low probability of belonging to random background fluctuations, thus ensuring that the labeled quantum dot-enriched micro-regions have statistically significant specific binding characteristics and effectively suppressing the generation of false positive micro-regions. The center coordinates and area of ​​each enriched micro-region are recorded, providing a data foundation for subsequent precise localization of terahertz spectroscopy and the introduction of micro-region area parameters into the feature vector.

[0071] Example 7, the method for acquiring terahertz time-domain spectral signals in step 4 is as follows: based on the determined coordinates of the enriched micro-regions, the electric sample stage is controlled to move each enriched micro-region sequentially to the focal position of the terahertz beam, and the terahertz time-domain spectral signals are acquired in transmission mode; the terahertz frequency range is 0.1THz-4.0THz, the terahertz spot diameter is 100μm-150μm, and each spectrum is acquired by 100-150 time-domain scans and the average value is taken; at the same time, the terahertz time-domain spectrum of the blank area without fluorescent label is acquired as a reference signal; the ambient temperature is controlled at 23±1 degrees Celsius and the relative humidity is below 5%.

[0072] By employing the above technical solution, terahertz spectral acquisition utilizes a micro-area-guided focusing mode, a key design feature for improving the signal-to-noise ratio. The terahertz spot diameter, 100-150 μm, is comparable to the typical quantum dot-labeled enrichment micro-area size in algal powder films, ensuring a high spatial match between the terahertz detection area and the positively labeled area, maximizing the acquisition of the specific terahertz response induced by the quantum dot-heavy metal complex. If a wide-beam, large-area irradiation is used, the matrix background signal from the unlabeled area dominates the total signal, diluting the specific signal from the positive area to an unrecognizable level. Each spectral acquisition involves 100-150 time-domain scans, with averaged results. Statistical averaging effectively suppresses random noise sources such as femtosecond laser pulse energy fluctuations, terahertz antenna noise, and environmental electromagnetic interference, improving the signal-to-noise ratio by approximately one order of magnitude. Maintaining an ambient temperature of 23℃±1℃ and a relative humidity of less than 5% is essential for ensuring the repeatability of terahertz spectra. Temperature fluctuations can cause response drift in optoelectronic devices and changes in the sample's refractive index. Water vapor has dense molecular rotation absorption lines in the terahertz band. If the relative humidity is higher than 5%, the water vapor absorption peaks will overlap with the sample signal, causing spectral distortion.

[0073] Example 8, step 5 of the multimodal fusion analysis includes the following sub-steps:

[0074] Step 51: Perform a fast Fourier transform on the terahertz time-domain spectral signal and the reference signal acquired in the previous step to calculate the terahertz absorption coefficient α(ν) and refractive index n(ν).

[0075] Step 52: For each quantum dot-labeled enriched micro-region i, calculate the terahertz absorption coefficient of the quantum dot-labeled region. Absorption coefficient relative to the unlabeled region of the same sample Differential absorption spectrum To eliminate the contribution of algal powder matrix background absorption;

[0076] Step 53, construct the joint feature vector , ;in These are discrete frequency points selected at equal intervals within the range of 0.1THz to 4.0THz, where N is the total number. It is the normalized fluorescence intensity of the arsenic-specific quantum dot in the i-th enriched micro-region; It is the normalized fluorescence intensity of the mercury-specific quantum dot in the i-th enriched micro-region; It is the area of ​​the i-th enriched micro-region.

[0077] By employing the above technical solution, differential absorption spectroscopy extraction is the core preprocessing step in multimodal fusion analysis. (Calculation) The physical basis is:

[0078] The components of algal powder matrix, such as proteins, polysaccharides, lipids, and pigments, exhibit a broadband, non-characteristic absorption background in the terahertz band. This background can be considered approximately consistent between the labeled and adjacent unlabeled regions of the same film. Subtracting the two, the background absorption of the algal powder matrix is ​​eliminated, and the residual signal in the differential spectrum mainly originates from the new terahertz absorption contribution introduced by the complex formed after the quantum dot-labeled probe specifically binds to heavy metal ions. The eigenvector simultaneously incorporates the terahertz differential absorption coefficient, the normalized fluorescence intensity of the two types of quantum dots, and the microarea area. The design intention is to achieve information complementarity: the terahertz spectral features mainly provide the vibrational mode fingerprint of the quantum dot-heavy metal complex and concentration-related absorption intensity information; the fluorescence intensity provides reference information on the probe binding amount; and the microarea area reflects the spatial aggregation degree of heavy metals in the sample. These three types of information are relatively independent in physical origin, and their fusion can significantly improve the accuracy of concentration inversion.

[0079] Example 9: The multi-task learning model based on causal inference in step 5 adopts a causal multi-task learning architecture, specifically including:

[0080] Construct a causal graph structure G=(V,E), where the node set V contains terahertz spectral feature nodes. Quantum dot fluorescence characteristic nodes Algal powder matrix confusion factor node Z, arsenic concentration node and mercury concentration nodes The edge set E is composed of causal relationships determined by domain knowledge;

[0081] A dual-channel convolutional neural network is used as the encoder for terahertz spectral features and quantum dot fluorescence features, and a fully connected regression head is used as the predictor for arsenic and mercury concentrations. A causal regularization term is introduced to eliminate the confounding effect of algal powder matrix confusion factors and ensure cross-batch generalization ability.

[0082] By adopting the above technical solution, the fundamental difference between the causal multi-task learning architecture and the traditional black-box multi-task learning model lies in that it explicitly models the causal interaction path of the algal powder matrix background as a confounding factor. In the causal graph structure, the algal powder matrix confounding factor node Z simultaneously points to terahertz spectral feature nodes. This is because differences in pigment types, protein content, and polysaccharide composition among algal powders from different origins and batches introduce significant baseline shifts and broadband absorption differences in the terahertz spectrum. If this obfuscation path is not eliminated, the model may learn a spurious correlation between matrix features and output concentration, performing well on training batches but experiencing a sharp decline in predictive ability when faced with new batches.

[0083] The expression for the loss function is:

[0084] ;

[0085] It is the model's total causal regularization loss. It is a prediction loss, using the mean squared error loss function. It is the Hilbert-Schmidt independence criterion term, used to constrain the algal powder matrix confusion factor Z to be independent of the predicted concentration C; It is the maximum mean difference term, used to eliminate differences between sample batches. , Domain offset between; , These are regularization hyperparameters, corresponding to the weight coefficients of the independent constraint term and the domain offset constraint term, respectively.

[0086] Introducing the Hilbert-Schmidt independence criterion forces Z to be statistically independent of the predicted concentration C, effectively cutting off the confusion path and making the mapping relationship from the learned features to the concentration causally significant. The maximum mean difference term further enhances the model's generalization ability to unseen batches of samples by minimizing the distribution difference between the training and test batches in the feature space. The dual-channel convolutional neural network, acting as the encoder, can automatically learn decoupled feature representations sensitive to arsenic and mercury respectively from high-dimensional terahertz spectra, avoiding the limitations of manually designed features.

[0087] Example 10, the specific method of step 6 is: to invert the predicted arsenic concentration value obtained in step 5. and predicted mercury concentration The values ​​are compared with the set limits for arsenic and mercury in algae and their products; the 95% confidence interval of the model prediction is estimated using the Bayesian dropout method; if the predicted concentration exceeds the corresponding limit, it is judged as exceeding the standard, otherwise it is judged as qualified, and the judgment conclusion, the predicted concentration value and its confidence interval are output together to generate a standardized test report.

[0088] By employing the above technical solution, the Bayesian dropout method is used to estimate the prediction confidence interval. The principle is to perform multiple random forward propagations with dropout during model inference, randomly discarding a portion of the network node outputs each time, generating a set of slightly different prediction values. The distribution of this set of prediction values ​​reflects the model's response to the uncertainty of the input data: if the features of a test sample are located in a densely distributed region of the training data, multiple predictions will be highly consistent, resulting in a narrow confidence interval; if the test sample deviates from the training distribution, the extrapolation results of different substructures of the network for that sample will differ significantly, widening the confidence interval and reflecting a decrease in prediction reliability. Setting a 95% confidence interval means that the probability of the true concentration value falling within this interval is 95%, providing a quantitative reliability reference for screening and judgment. Combining this uncertainty measure with the limit value comparison can avoid misjudgments due to small deviations in single-point prediction values, which is particularly important for judging samples near the boundary of the limit value.

[0089] The following describes the implementation principle of the present invention using specific embodiments:

[0090] A food and drug testing center received a batch of commercially available spirulina tablets. The product label indicated that the raw material was *Spirulina platensis*, the place of origin was Inner Mongolia Autonomous Region, China, and the production batch number was 20260315. The center was required to complete rapid screening for arsenic and mercury content and issue a test report within 4 hours. Six individually packaged tablets from this batch were sampled for testing.

[0091] Testing process:

[0092] Sample pretreatment: Take two spirulina tablets from each individual package, grind them into powder in an agate mortar, and pass them through an 80-mesh sieve. Weigh 100.0 mg of each sample powder, place them in a clean agate mortar, add 1.0 mL of high-purity anhydrous ethanol, and manually grind for 5 min until a uniform paste is formed. Coat the paste evenly onto a high-resistivity silicon wafer substrate with a thickness of 505 μm and a resistivity of 15 kΩ·cm. The average transmittance in the 0.1 THz to 4.0 THz frequency range was tested to be 68%. Spin coat the film at 800 rpm for 30 s. After film formation, measure the thickness at five points (four corners and center) on the film surface using a micrometer. The average thickness was 108 μm, with a deviation of no more than 12 μm at each point, meeting the requirement of 100 μm ± 20 μm. Dry the film in a nitrogen atmosphere at 40℃ for 15 min, then remove and seal it in a desiccator. Prepare six test films using the same method, numbered S01 to S06.

[0093] Probe Preparation: Preparation of arsenic-specific quantum dot-labeled probes. 50 mg of CdSe / ZnS core-shell quantum dots with an emission wavelength of 522 nm were dispersed in 10 mL of chloroform. Separately, 200 mg of glutathione was dissolved in 10 mL of methanol, and the pH was adjusted to 10 with NaOH. The two solutions were mixed and stirred at 60 °C for 12 h under nitrogen protection. Excess acetone was added to the reaction solution to precipitate the precipitate, which was collected by centrifugation at 8000 rpm for 5 min. This centrifugation was repeated twice. The precipitate was dispersed in 10 mL of phosphate buffer, and 20 mg each of EDC and NHS were added. The mixture was activated at room temperature for 30 min. 15 mg of the arsenic-specific recognition peptide Cys-Gly-Gly-Cys-Gly-Gly-Gly was added, and the reaction was stirred at 4 °C in the dark for 12 h. The reaction solution was purified three times by ultrafiltration using a 30 kDa ultrafiltration tube at 4000 rpm, with phosphate buffer added to the original volume each time. Finally, an arsenic-specific quantum dot-labeled probe stock solution was obtained and diluted to 100 nM with phosphate buffer for later use.

[0094] Preparation of mercury-specific quantum dot-labeled probes. 50 mg of CdSe / ZnS core-shell quantum dots with an emission wavelength of 618 nm were dispersed in 10 mL of chloroform. 200 mg of dihydrolipoic acid was dissolved in methanol and then added to the quantum dot chloroform solution. The mixture was stirred at 60 °C for 6 h. The reaction solution was precipitated with n-hexane, collected by centrifugation, and dispersed in Tris-HCl buffer. 10 mL of this solution was taken and thiol-modified T-rich single-stranded DNA solution was added to a molar ratio of 1:10. After incubation at room temperature for 12 h, an equimolar amount of mercury-specific quantum dot-labeled probe was added. 2+ Aptamer sequences were obtained and hybridized at 37°C for 2 hours. Mercury-specific quantum dot-labeled probe stock solution was obtained after ultrafiltration purification and diluted to 100 nM with Tris-HCl buffer for later use.

[0095] Labeling Incubation: Mix 0.5 mL each of arsenic-specific quantum dot labeled probe stock solution and mercury-specific quantum dot labeled probe stock solution, add 9 mL of phosphate buffer and 5 μL of Tween-20 to obtain a dual-channel quantum dot labeled probe mixture with a total quantum dot concentration of 10 nM and a Tween-20 volume fraction of 0.05%, for a total volume of 10 mL. Place the S01 to S06 test films flat on the support in the humidified chamber, and use a pipette to add 100 μL of the mixture evenly to each film surface. Seal the humidified chamber tightly and incubate at 37°C in the dark for 30 min. After removal, rinse the film surface three times with 1.0 mL of phosphate buffer each time, and dry under a nitrogen stream to obtain the S01 to S06 quantum dot labeled test films.

[0096] Dual-modal synchronous data acquisition: The S01 quantum dot-labeled test film was fixed on the sample stage of the measurement platform. The quantum microscopy imaging module was activated, and the 520nm channel fluorescence image was acquired by EMCCD with a 405nm laser excitation and an exposure time of 200ms. Then, the excitation was switched to 532nm laser excitation, and the 620nm channel fluorescence image was acquired with an exposure time of 200ms. The imaging field of view was 500μm × 500μm, and the spatial resolution was 0.8μm.

[0097] The two-channel images were pixel-level superimposed, and the mean μ and standard deviation σ of the full-field fluorescence background were calculated. The background mean μ = 127 (grayscale value, 0 to 255 for 8-bit images), and the standard deviation σ = 18. The threshold was set as μ + 3σ = 181. After adaptive thresholding, 14 quantum dot-labeled positive micro-regions were identified on the S01 film, numbered D01 to D14. The center coordinates and fluorescence intensity of each micro-region were recorded. Taking micro-region D03 as an example: coordinates x = 213 μm, y = 87 μm; fluorescence intensity I_As = 215 (normalized value 0.84) in the 520 nm channel, fluorescence intensity I_Hg = 198 (normalized value 0.78) in the 620 nm channel; micro-region area A = 2850 μm², equivalent diameter approximately 60 μm. The fluorescence intensity of both channels in this micro-region was significantly higher than the threshold, indicating the presence of positive signals for arsenic and mercury probe binding in this region.

[0098] The motorized sample stage was used to sequentially move 14 positive micro-regions to the focal point of the terahertz beam. The terahertz beam had a diameter of 120 μm, a frequency range of 0.1 THz to 4.0 THz, and a frequency resolution of 10 GHz. Each spectrum was scanned 120 times, and the time-domain waveforms were averaged. A blank region without fluorescent label was placed between every two positive micro-regions as a reference point, and the reference signal was acquired using the same method. The ambient temperature was maintained at 23.2℃, and nitrogen purging was used to maintain a relative humidity of 3.5% in the optical path.

[0099] The dual-mode synchronous data acquisition of films S02 to S06 was completed sequentially. The entire acquisition process took approximately 3.5 hours.

[0100] Multimodal fusion analysis and concentration inversion:

[0101] Fast Fourier Transform was performed on the time-domain signal and corresponding reference signal acquired for each positive micro-region. Using the average film thickness measurement value d = 108 μm, the absorption coefficients of the labeled and unlabeled regions of each micro-region were calculated. Taking the D03 microregion as an example, its main terahertz spectral data are as follows: at 1.68 THz, the absorption coefficient of the labeled region is... =47.6cm -1 Absorption coefficient of adjacent unlabeled regions =38.2cm -1 Differential absorption coefficient =9.4cm -1 At 2.35 THz, =7.8cm -1 Four identifiable differential absorption characteristic peaks were observed across the entire frequency band in the range of 0.8THz to 3.2THz, with peak positions at 1.12THz, 1.68THz, 2.35THz, and 2.91THz, respectively.

[0102] Construct the joint feature vector according to step 53. Construct feature vectors for each of the 14 microregions of the S01 thin film. , i=1 to 14. Summarize the joint feature vectors of all 83 positive microregions from the 6 films.

[0103] Eighty-three sets of joint feature vectors were input into the trained causal multi-task learning model. The model used a dual-channel convolutional neural network encoder to extract decoupled representations of terahertz spectral features and quantum dot fluorescence features, which were then output as predicted values ​​for arsenic and mercury concentrations via a fully connected regression head. During the training phase, the model's parameters were optimized using 36 sets of data from spirulina standard samples with known concentrations, and optimal selection was achieved through cross-validation based on the sample matrix type. , Regularization coefficient values, root mean square error of training set =0.032mg / kg, =0.018mg / kg.

[0104] Taking the D03 micro-region as an example, the model outputs a predicted arsenic concentration value. =0.78 mg / kg, predicted mercury concentration =0.12mg / kg.

[0105] The predicted values ​​of the 14 micro-regions were taken as an area-weighted average. The weighted average concentration of arsenic in the SO1 film was 0.67 mg / kg, and the weighted average concentration of mercury was 0.09 mg / kg.

[0106] Similarly, the weighted average concentrations of arsenic and mercury in films SO2 to SO6 were calculated. The concentration ranges for the six films were: arsenic 0.41 mg / kg to 0.78 mg / kg, and mercury 0.03 mg / kg to 0.15 mg / kg. The intra-batch coefficients of variation were 18.3% for arsenic and 22.7% for mercury, consistent with the uneven distribution of heavy metals in spirulina.

[0107] Results Judgment and Report Generation: The preset limit thresholds are based on the national food safety standards for algae products, which specify an inorganic arsenic limit of 0.5 mg / kg and a total mercury limit of 0.1 mg / kg. The arsenic detection value corresponds to inorganic arsenic plus trivalent methylarsenic. The inorganic arsenic limit is used for compliance determination, and the detection result is not equivalent to the total arsenic content of the sample. The mercury detection value only includes inorganic mercury and small-molecule methylmercury; large-molecule organic mercury cannot be detected. Screening is performed according to the national standard total mercury limit for risk screening. For samples exceeding the limit, accurate quantification of total mercury requires retesting using the national standard wet digestion instrumental method. The weighted average arsenic concentrations of films S01, S03, and D06 are 0.67 mg / kg, 0.72 mg / kg, and 0.78 mg / kg, respectively, exceeding the 0.5 mg / kg threshold. The weighted average mercury concentrations of films S01 and S05 are 0.09 mg / kg and 0.15 mg / kg, respectively, with the latter exceeding the 0.1 mg / kg threshold.

[0108] For each microregion's predicted value, Bayesian dropout inference was performed, with 100 random forward propagations, and a 95% confidence interval was calculated. Taking microregion D03 in S01 as an example, the 95% confidence interval for the predicted arsenic value of 0.78 mg / kg is [0.65, 0.91] mg / kg. The lower limit of 0.65 mg / kg is greater than the limit of 0.5 mg / kg, indicating that the determination of arsenic exceedance in this microregion has statistical confidence. In sample S02, the predicted arsenic value in a certain microregion is 0.52 mg / kg, with a 95% confidence interval of [0.41, 0.63] mg / kg. The lower limit is lower than the limit while the upper limit is higher than the limit, and the conclusion is suspected exceedance. It is recommended to verify using standard methods.

[0109] The test report includes information such as report number, sample name, production batch number, test date, test items, test methods, and limit standards. Figure 2 and Figure 3 As shown;

[0110] The test results are shown in Table 1:

[0111] Table 1

[0112] Sample number Predicted arsenic concentration (mg / kg) 95% confidence interval for arsenic (mg / kg) Predicted mercury concentration (mg / kg) 95% confidence interval for mercury (mg / kg) Judgment Conclusion S01 0.67 0.52 to 0.82 0.09 0.04 to 0.14 Arsenic exceeding the standard S02 0.41 0.28 to 0.54 0.03 0.01 to 0.05 qualified S03 0.72 0.58 to 0.86 0.08 0.03 to 0.13 Arsenic exceeding the standard S04 0.48 0.36 to 0.60 0.06 0.02 to 0.10 qualified S05 0.52 0.41 to 0.63 0.15 0.10 to 0.20 Mercury levels exceeded the standard, and arsenic levels were suspected to be excessive. S06 0.78 0.65 to 0.91 0.11 0.06 to 0.16 Arsenic levels exceeded the standard, and mercury levels were suspected to be excessive.

[0113] Spectral quality assessment: The average spectral signal-to-noise ratio of the six films was 31.2 dB, and the peak dynamic range of the terahertz time-domain signal in each micro-region was greater than 25 dB, indicating that the data quality met the screening requirements.

[0114] Judgment Conclusions: Of the six samples, S01, S03, and S06 were found to have excessive arsenic, S05 was found to have excessive mercury, and S02 was deemed acceptable. Although the upper limit of the 95% confidence interval for arsenic in S04 exceeded the limit, the center point (0.48 mg / kg) was below the inorganic arsenic limit of 0.5 mg / kg, thus it was deemed acceptable overall. The center point of S05 was above the limit and the interval crossed the limit, indicating suspected arsenic exceedance. The measured mercury in S06 was 0.11 mg / kg, with a range of 0.06-0.16 crossing the limit, indicating suspected mercury exceedance. The mercury in S05 was 0.15 mg / kg, and the lower confidence limit was equal to the limit of 0.1 mg / kg, indicating mercury exceedance. If total arsenic content of all components in the sample needs to be determined, a wet digestion pretreatment combined with the national standard instrumental method is required for retesting, and supplementary determination of pentavalent methylarsenic and arsenic sugar content is necessary.

[0115] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy, characterized in that, Includes the following steps: Step 1: Prepare a standardized algae powder test film from the algae powder sample; Step 2: Prepare arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes respectively; the arsenic-specific quantum dot labeled probe is composed of a first quantum dot nanocrystal core and an arsenic-specific recognition group; the mercury-specific quantum dot labeled probe is composed of a second quantum dot nanocrystal core and a mercury-specific recognition group. Step 3: Arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes are directionally labeled and incubated with the algae powder test film, so that the quantum dot labeled probes specifically bind to arsenic and mercury in the algae powder sample. After washing and drying, the quantum dot labeled algae powder test film is obtained. Step 4: Perform dual-modal synchronous data acquisition on the quantum dot-labeled algal powder test film on the same measurement platform, including: acquiring fluorescence microscopic images of the quantum dot-labeled probe in quantum microscopic imaging mode to locate the quantum dot-labeled enriched micro-regions; and acquiring terahertz time-domain spectral signals of each enriched micro-region in terahertz time-domain spectral mode, guided by the located enriched micro-regions. Step 5: Perform multimodal fusion analysis on the collected dual-modal data, including: extracting the differential terahertz absorption spectrum of the quantum dot labeled region relative to the unlabeled region, constructing a joint feature vector that integrates terahertz spectral features and quantum micro-fluorescence features, and simultaneously inverting arsenic and mercury concentrations based on a causal inference multi-task learning model. Step 6: Compare the obtained arsenic and mercury concentrations with the preset limit thresholds and output the screening judgment results.

2. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 1, characterized in that, Step 1 includes the following sub-steps: Step 11: Weigh 100±0.1mg of algae powder sample, add high-purity anhydrous ethanol as a dispersion medium, grind into a uniform paste, and then uniformly coat it onto a high-resistivity silicon wafer substrate. Step 12: Use a spin coater to spin coat algae powder at a speed of 700rpm-900rpm to form a film with a thickness of 100±20μm; Step 13: Dry the film under a nitrogen atmosphere at 38℃-42℃ for 15-18 minutes to obtain the algae powder test film; the thickness of the high-resistivity silicon wafer substrate is 500±10μm, the resistivity is greater than 10kΩ·cm, and the transmittance in the 0.1-4.0THz frequency band is not less than 60%.

3. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 2, characterized in that, The preparation method of the arsenic-specific quantum dot-labeled probe in step 2 is as follows: CdSe / ZnS core-shell quantum dots with an emission wavelength of 520±5 nm and a quantum yield greater than 60% are selected as the core of the first quantum dot nanocrystal. Glutathione molecules are modified on its surface via ligand exchange. After activation with 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide, an arsenic-specific recognition peptide is coupled. The amino acid sequence of the arsenic-specific recognition peptide is Cys-Gly-Gly-Cys-Gly-Gly. After ultrafiltration purification, the arsenic-specific quantum dot-labeled probe is obtained. The glutathione bonded to the surface of the algal powder test film probe is in situ reduced to As(V) in an incubation environment. 3+ Then, As is chelated via peptide dithiol coordination. 3+ Trivalent methylarsenic has no specific binding ability to pentavalent methylarsenic or arsenose.

4. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 3, characterized in that, The preparation method of mercury-specific quantum dot-labeled probe in step 2 is as follows: CdSe / ZnS core-shell quantum dots with an emission wavelength of 620±5nm and a quantum yield greater than 55% are selected as the core of the second quantum dot nanocrystal. They are converted into water-soluble quantum dots modified with dihydrolipoic acid by ligand exchange method. Then, a thymine-rich single-stranded DNA sequence (5'-SH-(CH2)6-TTTTTTTTTTTTTTTT-3') is grafted onto its surface via Au-S bonds, and Hg is added. 2+ Hybridization was performed using the aptamer sequence (5'-TTCTTTCTTCCCCTTGTTTGTT-3') based on T-Hg. 2+ -T coordination chemistry mechanism enables specific recognition of mercury ions.

5. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 4, characterized in that, Step 3 includes the following sub-steps: Step 31: Mix arsenic-specific quantum dot labeled probes and mercury-specific quantum dot labeled probes at a molar ratio of 1:1, dilute with phosphate buffer to a total quantum dot concentration of 10 nM, and add 0.05% Tween-20 surfactant to obtain a dual-channel quantum dot labeled probe mixture. Step 32: 100 μL of the dual-channel quantum dot labeled probe mixture is uniformly dropped onto the surface of the algae powder test film and incubated in a humidity-saturated sealed container at 36-40°C in the dark for 28-35 minutes. Step 33: After incubation, rinse the film surface with phosphate buffer to remove unbound probes, and dry it under nitrogen flow to obtain a quantum dot labeled algae powder test film.

6. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 5, characterized in that, The method for locating the enriched micro-regions in step 4 is as follows: An arsenic-specific quantum dot-labeled probe was excited with a 405 nm laser, and fluorescence microscopy images were acquired in the 520 nm channel. A mercury-specific quantum dot-labeled probe was excited with a 532 nm laser, and fluorescence microscopy images were acquired in the 620 nm channel. The imaging field of view was 500 μm × 500 μm. Pixel-level superposition analysis was performed on the two fluorescence images. An adaptive threshold segmentation algorithm was used to identify pixel regions with fluorescence intensity higher than three times the background mean, which were marked as quantum dot-labeled enriched microregions. The center coordinates and corresponding fluorescence intensity values ​​of each enriched microregion were recorded. The quantum dot-labeled enriched microregions represent the potential arsenic / mercury presence areas located by the specific binding of the quantum dot probe.

7. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 6, characterized in that, The acquisition method for the terahertz time-domain spectral signal in step 4 is as follows: Based on the determined coordinates of the enriched micro-regions, the electric sample stage is controlled to move each enriched micro-region sequentially to the focal position of the terahertz beam, and the terahertz time-domain spectral signal is acquired in transmission mode; the terahertz frequency range is 0.1THz-4.0THz, the terahertz spot diameter is 100μm-150μm, and each spectrum is acquired by 100-150 time-domain scans and the average value is taken; at the same time, the terahertz time-domain spectrum of the blank area without fluorescent label is acquired as a reference signal; the ambient temperature is controlled at 23±1 degrees Celsius and the relative humidity is below 5%.

8. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 7, characterized in that, Step 5, the multimodal fusion analysis, includes the following sub-steps: Step 51: Perform a fast Fourier transform on the terahertz time-domain spectral signal and the reference signal acquired in the previous step to calculate the terahertz absorption coefficient α(ν) and refractive index n(ν). Step 52: For each quantum dot-labeled enriched micro-region i, calculate the terahertz absorption coefficient of the quantum dot-labeled region. Absorption coefficient relative to the unlabeled region of the same sample Differential absorption spectrum To eliminate the contribution of algal powder matrix background absorption; Step 53, construct the joint feature vector , ;in These are discrete frequency points selected at equal intervals within the range of 0.1THz to 4.0THz, where N is the total number. It is the normalized fluorescence intensity of the arsenic-specific quantum dot in the i-th enriched micro-region; It is the normalized fluorescence intensity of the mercury-specific quantum dot in the i-th enriched micro-region; It is the area of ​​the i-th enriched micro-region.

9. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 8, characterized in that, Step 5, the multi-task learning model based on causal inference, adopts a causal multi-task learning architecture, specifically including: Construct a causal graph structure G=(V,E), where the node set V contains terahertz spectral feature nodes. Quantum dot fluorescence characteristic nodes Algal powder matrix confusion factor node Z, arsenic concentration node and mercury concentration nodes The edge set E is composed of causal relationships determined by domain knowledge; A dual-channel convolutional neural network is used as the encoder for terahertz spectral features and quantum dot fluorescence features, and a fully connected regression head is used as the predictor for arsenic and mercury concentrations. A causal regularization term is introduced to eliminate the confounding effect of algal powder matrix confusion factors and ensure cross-batch generalization ability.

10. The rapid screening method for arsenic and mercury in algal powder based on terahertz time-domain spectroscopy and quantum microscopy according to claim 9, characterized in that, The specific method for step 6 is: to invert the predicted arsenic concentration value obtained in step 5. and predicted mercury concentration The values ​​are compared with the set limits for arsenic and mercury in algae and their products; the 95% confidence interval of the model prediction is estimated using the Bayesian dropout method; if the predicted concentration exceeds the corresponding limit, it is judged as exceeding the standard, otherwise it is judged as qualified, and the judgment conclusion, the predicted concentration value and its confidence interval are output together to generate a standardized test report.

Images